Computational Nuclear Astrophysics

Page created by Calvin Santiago
 
CONTINUE READING
Computational Nuclear Astrophysics
Computational Nuclear Astrophysics
Computational Nuclear Astrophysics
Key Science Drivers of Computational Nuclear
                       Astrophysics

 Primary Goal: Explanation of the Origin of the Elements and Isotopes
 Overwhelmingly, Elements are produced in Stars - quiescently or explosively
 Core-Collapse Supernovae (CCSN) - the Deaths of Massive Stars and Birth of
  Neutron Stars
 Thermonuclear Supernovae - the Source of much of the Iron Peak
 Novae - source of some light elements
 X-ray bursts - the rp-Process Nuclei
 Merging Neutron stars - with CCSN, the likely source of the r-process Nuclei
 Stellar Evolution involves nuclear reaction rates generated theoretically or
  experimentally - convective processes and magnetic couplings - multi-dimensional
 Stellar Explosions are always Multi-dimensional, requiring state-of-the-art
  radiation/hydrodynamic simulations with significant Nuclear Physics input.

 Nuclear astrophysics entails sophisticated multi-dimensional numerical simulations
  employing the latest computational tools and the most powerful supercomputers of
  the DOE complex to address key goals of the Office of Nuclear Physics.
Computational Nuclear Astrophysics
Nuclear Processes in the Cosmos
                                                        FRIB

                                            (by 2019 early comp)

               Swift

Chandra

                                                  NuStar
Computational Nuclear Astrophysics
The Cosmic Laboratory;
 Understanding nuclear processes at
 the extreme temperature & density
 conditions of stellar environments!

Stellar matter
Stellar explosions
White Dwarf matter
Neutron Star matter
Quark Star matter
Big Bang

Field requires close communication between
nuclear experimentalists, theorists, stellar
modelers and stellar observers (astronomers)
Computational Nuclear Astrophysics
Core-Collapse
 Supernova
 Explosions
 A 7(+) dimensional problem;
        Nuclear EOS;
       Nucleosynthesis
Computational Nuclear Astrophysics
Partnership between Nuclear Astrophysicists and
                       Applied Mathematicians to Create
                   State-of-the-Art Computational Capabilities

   2nd-order, Eulerian, unsplit, compressible hydro
   PPM and piecewise-linear methodologies
   Multi-grid Poisson solver for gravity
   Multi-component advection scheme with reactions
   Adaptive Mesh Refinement (AMR) - flow control, memory management, grid
    generation
   Block-structured hierarchical grids
   Subcycles in time (multiple timestepping - coarse, fine)
   Sophisticated synchronization algorithm
   BoxLib software infrastructure, with functionality for serial distributed and shared
    memory architectures
   1D (cartestian, cylindrical, spherical); 2D (Cartesian, cylindrical); 3D (Cartesian)
   Multigroup Transport with v/c terms and inelastic scattering
   Uses scalable linear solvers (e.g., hypre) with high-performance preconditioners that
    feature parallel multi-grid and Krylov-based iterative methods - challenging!
   Example partnership: John Bell, Ann Almgren, Weiqun Zhang, Louis Howell, Adam
    Burrows, Jason Nordhaus - LBNL, LLNL, Princeton
Computational Nuclear Astrophysics
2D:2.3

“Inverse” Energy
Cascade in 2D -
 Buoyancy-
Driven
Convection has
(anomalously) a
lot of large-scale
power - Often
confused for the
SASI
Computational Nuclear Astrophysics
Character of 3D turbulence and Explosion Very
          Different from those in 2D
Computational Nuclear Astrophysics
Computational Nuclear Astrophysics
Buoyancy-driven Bubbles!
Magnetic CCSN Explosions: Might be the Source of Some R-Process
Multi-group “2 1/2”-D
  Radiation-Magneto-
Hydrodynamic (RMHD)
 simulations of Core-
 Collapse Supernovae
Sample Computational Requirements for Future Core-
                 Collapse Supernova Simulations

Platform Space             Neutrino #fν        Matrix   Ops./Δt

Current       256x32x64    8x12x14    20 GB    2 TB     6x1012

Near-         512x64x128   12x24x20   600 GB   200 TB   2x1015
Term

Exa-Scale 512x128x256 24x24x24        6 TB     3 PB     8x1016

“Full         512x128x256 24x24x24    6 TB     80 PB    4x1019
Coupling”

 Kotake et al. 2012
Cycle and Memory Requirements for Supernova Simulations

   1985 (1D)          - ~102-3 CPU-hours per run; 10 Gbytes memory
   1995 (low 2D)      - ~105-6 CPU-hours per run; 100 Gbytes memory
   2005 (medium 2D)    - ~106 CPU-hours per run; 102 cores; Tbytes memory
   2010 (low 3D)      - ~106-7 CPU-hours per run; 103-4 cores; Tbytes
    memory
   2015 (medium 3D)   - ~107-8 CPU-hours per run; 105 cores; 0.2-1 Pbytes
    memory
   2020 (heroic 3D)   - ~108-9 CPU-hours per run; 105-6 cores; >10 Pbytes
    memory
Short-Hard GRB Model: Merger of Neutron Stars -
             Site of the R-Process?
The R-Process:
 Core-Collapse Supernovae,
 Merging Neutron Stars, or
Hypernovae - Multiple Sites?
Reach of FRIB - Link toGoals   far off Stability
                       Nuclear Astrophysics

               Nuclear Masses & decay properties
               Neutron halos
               Disappearance of shell structure
               Emergence of new shapes,
               New collective modes of excitation
               Mapping the driplines
               Islands of stability

                   FRIB reach
R-Process Nucleosynthesis
Neutrino Oscillations
in Core-Collapse Supernovae:
 A Computational Challenge
Neutrino Oscillations and Self-Coupling-
       Coherent Neutrino Flavor Evolution
   Wigner density matrix, ensemble-averaging

                   Diagonal elements: real numbers:
                   Phase-Space densities
                   Off-diagonal elements: complex
                   numbers: Macroscopic Overlap
                   densities
                                                  (Real part)

                                                  (Imaginary part)
Neutrino Oscillations and Collective Self-
                      interactions

6 species x 6 species = 36
transport densities/variables!
Type Ia
(Thermonuclear)
   Supernova
   Explosions
Turbulent Thermonuclear Flame Front
X-Ray Bursts - The
   rp-Process
Understanding the X-ray sky: rp-process in X-ray bursts
                            Multi-D X-ray burst simulations:
                             Link observables with FRIB
A. Spitkovsky           science to address open questions

                                                               FRIB reach for rate
                                                               measurements
                                                               with ReA3 + SECAR

                                                   100s statistical model reaction rates
     ab-initio or microscopic rates                (include α-cluster structure)
      Reliable rates were possible
                                                    Complement FRIB data set
      Inform statistical model
                                                    Astrophysical corrections to rates
X-Ray Burst rp-Process Nucleosynthesis
X-Ray Burst rp-Process Nucleosynthesis
X-Ray Burst rp-Process Nucleosynthesis
X-Ray Burst rp-Process Nucleosynthesis
X-Ray Burst rp-Process Nucleosynthesis
Nuclear Reaction Rate
                  Calculations
     General mathematical formalism for characterizing low-energy
     reaction cross sections

     Extrapolate cross sections to nearby energies but REQUIRES
     experimental data for constraint

    Most straightforward for Compound Nucleus type reactions

    May be applied to other reaction types using approximations:

a)     Radiative Capture
b)     Beta-delayed particle emission
15N(p, γ)16O
  12C(α,α) 12C

  12C(α,α) 12C

  12C(α,p)15N

  12C(α,γ)16O

  15N(p,p)15N

  15N(p,α)12C

  15N(p,α)12C

Multiple Channel Example
Nuclear Rate Calculations

  Thermonuclear reaction rates are an essential
       ingredient for any stellar model.

     A major obstacle in providing defensible
 uncertainties is that the rates are highly complex
  quantities derived from a multitude of nuclear
       properties extracted from laboratory
                   measurements.
   A solution to this challenge, devised recently by
Iliadis and collaborators, is STARLIB which contains
 Monte Carlo sampled probability densities of each
               rate at each temperature.
The MESA Stellar Evolution Code currently has over 400
registered users across the globe, with many users at
D.O.E nuclear physics sponsored programs.

MESA employs modern numerical approaches and is
written with present and future shared-memory, multi-
core, multi-thread and possibly hybrid architectures.
Computational Nuclear Astrophysics Addresses Key Experimental
    and Theoretical Goals of the Office of Nuclear Physics

   Computational nuclear astrophysics also supports important components of the
    DOE Office of Nuclear Physics Experimental program:

       The astrophysics of neutron-rich nuclei is one of four scientific ``legs" of the Facility for
        Rare Isotope Beams (FRIB).

       Supernova modeling efforts are important to the DOE's experimental Neutrino Physics
        program. The flagship experimental program in DOE's Intensity Frontier program includes
        a megadetector that could follow neutrino light curve of a CCSN

       The Nuclear equation of state is a third intersection with the DOE experimental program:
        JLab measurements constrain the nuclear symmetry energy, and, thus, the EOS for
        neutron-rich matter

   Nuclear astrophysics entails sophisticated multi-dimensional numerical simulations
    employing the latest computational tools and the most powerful supercomputers of
    the DOE complex to address key goals of the Office of Nuclear Physics.
Sample Computational Requirements for Future Core-
                 Collapse Supernova Simulations

Platform Space             Neutrino #fν        Matrix   Ops./Δt

Current       256x32x64    8x12x14    20 GB    2 TB     6x1012

Near-         512x64x128   12x24x20   600 GB   200 TB   2x1015
Term

Exa-Scale 512x128x256 24x24x24        6 TB     3 PB     8x1016

“Full         512x128x256 24x24x24    6 TB     80 PB    4x1019
Coupling”

 Kotake et al. 2012
Cycle and Memory Requirements for Supernova Simulations

   1985 (1D)          - ~102-3 CPU-hours per run; 10 Gbytes memory
   1995 (low 2D)      - ~105-6 CPU-hours per run; 100 Gbytes memory
   2005 (medium 2D)    - ~106 CPU-hours per run; 102 cores; Tbytes memory
   2010 (low 3D)      - ~106-7 CPU-hours per run; 103-4 cores; Tbytes
    memory
   2015 (medium 3D)   - ~107-8 CPU-hours per run; 105 cores; 0.2-1 Pbytes
    memory
   2020 (heroic 3D)   - ~108-9 CPU-hours per run; 105-6 cores; >10 Pbytes
    memory
R-Process Nucleosynthesis

Compare
Comparecalculated
        calculatedresults
                     resultswith
                            withabundance
                                  abundanceobservations
                                              observations??
Masses,
 Masses,half-lives,
          half-lives,n-capture
                      n-capturerates
                                 ratesof
                                       ofvery
                                          veryunstable,
                                               unstable,exotic
                                                         exoticnuclei
                                                                nucleineed
                                                                       needto
                                                                            tobe
                                                                               beknown
                                                                                  known
 Need experiments   and  nuclear  theory
 Need experiments and nuclear theory
Lagrangian Particle Advection Through the Shock -
   Heating Distribution in 3D - Early advection
Lagrangian Particle Advection Through the Shock -
                      Heating Distribution in 3D

FRAME 3:
Type Ia Supernova Facts
   Thermonuclear Explosion of the entire accreting C/O
    White Dwarf; Explosion lasts ~1 second
   Emits mostly Optical and Infrared light
   Used as a primary yardstick for the Cosmology. Can be
    seen across the Universe: Indicates the Universe is
    Accelerating - Nobel Prize
   Significant element production and ejection: Iron
    (radioactive Nickel), Ca, Si, S, Ar, …
   Light lasts months; Peak Luminosity ~ 1021 Megatonnes
    of TNT/second (very bright)
   Complete disassembly; Energy > 1028 Mtonnes of TNT
The Progenitors of Type Ia supernovae, whose use as an empirical tool
won this year’s Nobel Prize, are unknown. Identifying the nucleosynthetic
signatures in their spectra of different progenitors channels is needed to
reduce the uncertainty in the distance measurements, and quantify the
nature of the dark energy with increased precision.
X-Ray Burst rp-Process Nucleosynthesis
You can also read